AI-Driven Cost Estimation in Shipbuilding: A Hybrid Framework Integrating PSO and XGBoost

Abstract

Accurate shipbuilding cost estimation is important for project quotation, budget preparation, cost control, and operational decision-making. This study proposes an AI-driven hybrid framework for shipbuilding cost estimation by integrating particle swarm optimization (PSO) with the extreme gradient boosting (XGBoost) algorithm. The empirical dataset consists of 51 conventional vessels constructed by a shipyard between 2020 and 2024. Based on the characteristics of these vessels and their construction environment, 11 internal and external cost-influencing factors were selected as input variables. The dataset was divided into a training set and a test set at a ratio of 85% to 15%. The PSO-XGBoost model was compared with the conventional XGBoost model using R2 and MAPE. On the test set, the proposed model achieved an R2 of 0.8718 and a MAPE of 3.87%, compared with 0.8474 and 4.15%, respectively, for the conventional XGBoost model. SHAP analysis was further employed to examine the direction and relative contribution of the influencing factors, thereby improving the interpretability of the estimation results. Case-based validation using different conventional ship types further demonstrates the practical feasibility of the proposed framework. The findings indicate that PSO-based hyperparameter optimization can improve the predictive performance of XGBoost and provide interpretable decision support for shipbuilding cost management.